IVR 会誌　Jpn J Intervent Radiol Vol.20 No.4 2005

State of the Art The Statistics IVRists Should Know 1. Critical Knowledge before Needed Statistics to Conduct Clinical Research Department of Diagnostic Radiology, National Cancer Center Yasuaki Arai Statistics and Cancer Control Division, Research Center for Cancer Prevention and Screening, National Cancer Center Seiichiro Yamamoto Abstract Statistics is one of the most important tools needed to solve out clinical questions and to evaluate the correlation of various clinical phenomena. However, statistics cannot answer clinically valuable questions by itself. Only when statistics is applied with the scientifically appropriate methods and procedures does, it help us to clarify new thoughts and knowledge. This is called clinical epidemiology or clinical research methodology. Therefore, clinical researchers are strongly advised to learn them in addition to statistics. Such knowledge also helps clinical researchers to collaborate with statisticians. Key words ● Statistics ● Clinical epidemiology ● Clinical research methodology

2. What Provide by Statistical Analyses, Its Possibility and Danger Department of Environmetrics and Biometrics, Research Institute for Radiation Biology and Medicine, Hiroshima University Megu Ohtaki Abstract With the recent rapid improvement of performance of computer and network environment, various sophisticated statistical theories that were“castles in the air”can now be applied to data analysis easily. In view of such a background, the purpose of this paper is to outline several useful statistical methods such as summarization of data using classical descriptive statistics, correlation analysis, regression analysis and the latest methods called“computer intensive techniques”. It is also discussed how to apply them and how to interpret the results with possible pitfalls that should be avoided. Key words ● Outline of statistical methods ● Pitfall of analysis ● Statistical data analysis

3. Basic Biostatistical Terms in Interpreting Diagnostic Information and Outcomes Department of Radiology and Evidence Based Laboratory Medicine1), Kobe Graduate School of Medicine Yasushi Kaji, Daisuke Sugiyama1), Kunihiro Nishimura1), Ryo Sugihara, Kazuro Sugimura Abstract Recently, healthcare decisions have become complex because of many test results, new treatments, ethical problems, and so on. Biostatistics provides a systematic way to think about and analyze observational data in chosen populations. In this article, we summarize the basic biostatistical terms. We hope this article will promote a real discussion of the risks and benefits of the mana-gement option. Key words ● Biostatistics ● Probability ● Outcomes

4. Survival Analysis Department of Medicine, Kanagawa Dental College Toshio Morizane Abstract Survival is the most important outcome for patients with cancer and other diseases. The effect of an intervention on survival rate can be investigated using a randomized controlled trial. With the Kaplan-Meier method, the most widely used method of survival analysis in medicine, we can address censored cases that include drop-outs and those cases for which the outcome is unknown at the time of data analysis. In Kaplan-Meier survival analyses, time is divided into intervals marked by the time when the outcome occurs, and survival rates are calculated in each interval. The censored cases in each interval are included in the denominator. Survival rates are multiplied by the number of intervals to the interval to get the cumulative survival rates in each interval, producing a cumulative survival curve. In most cases the survival curves fit an exponential function in which a hazard rate is a determinant of the function. The survival curve shows the probability of outcome occurrence as a function of time and the average period to the outcome occurrence where the outcome has occurred in 50% of the cases. Average duration of life gained (ADLG), a statistical measure that is easy to understand, compares life-expectancy given a particular intervention with life-expectancy given no intervention. Two survival curves can be compared with the log-rank test or the generalized Wilcoxon test. When more than two survival curves are compared, the alpha level should be lowered. Key words ● Survival analysis ● Kaplan-Meier method ● Survival curve

5. Meta-analysis Department of Medicine, Kanagawa Dental College Toshio Morizane Abstract The statistical measures that quantify the efficacy of an intervention or the effect of a risk factor are the risk ratio, the relative risk reduction, the absolute risk reduction, the number needed to treat, and the odds ratio. The effect measures obtained in a single study follow characteristic distributions. For example, the natural logarithm of risk ratio follows the normal distribution. The standard deviation depends on the sample size: the larger the sample size, the smaller the standard deviation. In meta-analysis, effect measures are weighted (depending on the sample size of each study) and then synthesized. If the study subjects of each study are random samples drawn from a population and if covariates such as age, sex, disease severity, and other factors which might affect the outcome are the same, then the variation of the effect measure among studies is derived solely by chance. In this case the fixed effect model is appropriate. However, in many cases covariates and/or publication bias cause variation in the effect measure. The random effect models can be applied to such cases. The random effect models incorporate between-study variation for calculating the synthesized value of effect measures. The hierarchical Bayesian models address these variations in a more flexible way. When meta-analysis studies are evaluated, funnel plat, the L’Abbe graph, and the baseline graph should be read carefully. Key words ● Meta-analysis ● Medical statisfics ● Effect measure

6. Application of Multivariate Analysis Based on Clinical Example Department of Gastroenterology, Ogaki Municipal Hospital Takashi Kumada, Hidenori Toyoda, Seiki Kiriyama, Yasuhiro Sone, Makoto Tanikawa, Yasuhiro Hisanaga Abstract Remarkable development of personal computers and statistical analysis software have made it possible for clinicians to comparatively easily do multivariate analysis. We explained logistic regression model and Cox proportional hazard model showing clinical examples used well in the medical field. Logistic regression model was used for multivariate analysis of cases without censored cases. In this article we analyzed the factors to predict the effectiveness of interferon therapy in patients with chronic hepatitis C by using a logistic regression model. On the other hand, we used a Cox proportional hazard model which could process censored examples as multivariate analysis for life analysis. Cox proportional hazard model was used for evaluating factors associated with prognosis in patients with hepatocellular carcinoma. The ability to select statistical analysis depending on the purpose of use precisely is needed. Key words ● Cox proportional hazard model ● Logistic regression model ● Hazard ratio ● Odds ratio

Original Article Effectiveness of Preoperative Embolization in Surgical Reduction for Diffuse Plexiform Neurofibroma : Initial Experiences in Four Cases Department of Radiology, Saitama Medical School Junji Tanaka, Naoko Nishi, Masayuki Yuasa Abstract 　Cutaneous diffuse plexiform neurofibroma of neurofibromatosis type-1 sometimes grows to an enormous size and occasionally becomes malignant. It is preferable to remove it to the extent possible. However, it is usually very difficult to stop bleeding during surgery due to the abundance of vulnerable abnormal vessels within the tumor. We report here four cases of diffuse plexiform neurofibromas of the skin which were safely and almost totally removed by tumor reduction surgery with less than 20p of total bleeding using preoperative selective transarterial embolization. This standard technique has never been applied to cutaneous diffuse plexiform neurofibroma as pre-operative therapy. This may be a preferable option for patients with cutaneous plexiform neurofibroma when surgical reduction is planned. Key words ●Neurofibromatosis type-1 ●Plexiform neurofibroma ●Preoperative embolization therapy

Original Article Usefulness of Local Thrombolysis for the Treatment of Ileofemoral Deep Venous Thrombosis : A Comparison with to Systemic Intravenous Thrombolysis Department of Radiology, Oita University Faculty of Medicine Ryo Takaji, Hiro Kiyosue, Rieko Shuto, Masahisa Takuma, Syuichi Tanoue Mika Okahara, Hiromu Mori Abstract Purpose : To evaluate the usefulness of local thrombolysis for ileofemoral deep venous thrombosis (DVT) in comparison with systemic intravenous thrombolysis. Patients and Methods : Twenty-two consecutive patients with ileofemoral DVT were included in this comparative study. All but one patient underwent placement of an inferior vena cava filter before the treatments. Treatment method of either local thrombolysis (local thrombolysis group: n=10) or systemic thrombolysis (systemic thrombolysis group : n=12) was randomly selected for each patient. For the technique of local thrombolysis, a 4F catheter with multiple side holes was percutaneously placed into the thrombi via the ipsilateral femoral or popliteal vein. Continuous infusion of urokinase was performed via the catheter. The following items were assessed in both groups: total dosage of urokinase used, change in size of thrombi, change of symptoms, duration to the disappearance of symptoms, and complications. Follow-up CT was performed one to three weeks after the initiation of the treatments. Results : Mean dosage of urokinase used were 168×104U (range : 112～672×104U) in the local thrombolysis group and 114×104U (range : 72～540×104U) in the systemic thrombolysis group. In the local thrombolysis group, follow-up CT showed complete or nearly complete resolution of thrombi in all patients. Follow-up CT showed no change of the thrombi in 5 patients in the systemic group. Mean duration to disappearance of symptoms was 9.5 days (range : 1～16 days) in the local thrombolysis group and 12.5 days (range : 10～60 days) the in systemic thrombolysis group. No complications related to the procedures were observed. Conclusion : This study demonstrated the superiority of local thrombolysis to systemic thrombolysis with respect to thrombolytic effect and relief of symptoms. Local thrombolysis seems to be a safe and effective method for the treatment of ileofemoral DVT. Key words ●Deep venous thrombosis ●Catheter ●Thrombolysis ●Inferior vena cava filter

Case Report Two Cases of Acute Superior Mesenteric Artery Embolism Which Were Treated Successfully by Transcatheter Thromboaspiration Department of Radiology and Surgery1), Kyoto City Hospital Masto Tanikake*, Katsumi Hayakawa, Hirofumi Itou, Toshiki Shiozaki Hiroyuki Ueda, Kouichi Matsuo1), Eiji Yamamoto1), Taisuke Morimoto1) (*Department of Radiology, Osaka Medical College） Abstract 　We report two cases of acute superior mesenteric artery (SMA) embolism that were successfully treated with a combination of local thrombolysis, transcatheter thromboaspiration, and continuous arterial infusion. Although these two cases underwent local thrombolysis with urokinase, the thrombi were not dissolved but remained in the artery. We succeeded in removing the remaining thrombus by aspiration with an aspiration catheter (TVACTM) in one case and using a large-diameter guiding catheter in the other. Bowel resection could be avoided in both cases. Vascular intervention for acute SMA embolism is indicated only when the ischemic intestines are in a viable state. Usually, thrombolysis is selected first. If the thrombi cannot be cleared by this method, transcatheter thromboaspiration is considered to be useful for removal of remaining thrombi. Key words ●Embolism ●Mesentery ●Interventional procedure ●Catheter and catheterization

このページの先頭へ戻る